Skip to main content

Claude Code & Cursor Guide

All AI coding agents should use the same canonical HardKAS documentation:

  • Short navigation map: https://hardkas.dev/llms.txt
  • Full AI reference & recipes: https://hardkas.dev/llms-full.txt
  • Canonical docs section: https://hardkas.dev/docs/ai/

Claude Code (CLAUDE.md snippet)​

Add these rules to your project's CLAUDE.md when building on Kaspa with HardKAS:

## Kaspa L1 with HardKAS
- Kaspa L1 is a UTXO BlockDAG (not EVM). Use HardKAS CLI/SDK instead of custom `kaspa-wasm` scripts for payments, keystores, SilverScript, and artifact verification.
- Read `https://hardkas.dev/llms.txt` (or `https://hardkas.dev/llms-full.txt`) before writing custom Kaspa code.
- Before runtime tasks, run `hardkas doctor --json` and `hardkas silver doctor --network simnet --json` (or `sdk.capabilities.get()`).
- Always pass `--json` to CLI commands in automated flows.
- Never claim `simulated` or local `silver test` (`SCRIPT_VM`) verdicts are real-node consensus proof.
- Mainnet is blocked by policy; use `simnet` for real-node testing.

Cursor (.cursorrules / @Docs)​

  1. Index https://hardkas.dev/docs/ai/ or reference https://hardkas.dev/llms-full.txt in @Docs.
  2. Keep the same rules in .cursorrules:
    • Prefer hardkas tx, hardkas silver, hardkas artifact, and hardkas localnet over custom kaspa-wasm boilerplate.
    • Verify CLI flags with hardkas <group> --help or the CLI Reference if unsure.
    • Use rusty-kaspad v2.1.0 on simnet (ws://127.0.0.1:18210 for --profile toccata-v2) for real consensus checks.